McPherson AI
Observa private beta now open

Agent says it worked. Observa checks reality.

Observa is the governance and independent verification layer for action-taking AI agents. It maps what your agents can do, captures reviewable evidence, and checks supported workflow outcomes against the systems that hold the real state.

SHADOW mode first. No production control. Observa observes, evaluates, maps, and verifies without taking control of production actions.

Request Private Beta Access Watch the Demo Explore Observa →
Private beta open · n8n and OpenClaw SHADOW_ONLY · Authority NONE · Enforcement OFF
Public OpenClaw plugin on GitHub Public OpenClaw plugin on ClawHub
See Observa in action

Watch an AI agent move from runtime activity to reviewable governance evidence.

Observa captures what happened, keeps runtime claims separate from verified outcomes, and shows what governance would have done while remaining in SHADOW.

SHADOW ONLY · AUTHORITY NONE · ENFORCEMENT OFF

Product walkthrough

From a real agent task to a reviewable governance record.

An OpenClaw agent gets a real task and reaches outside the model. Observa discovers the real capability surface, a human approves the mapping and classifies it, and the boundary becomes a lifecycle record.

Observa product demo · 43 seconds · has sound; nothing plays until you press play. An agent gets a real task and reaches outside the model, Observa discovers the real capability surface, and a human approves and classifies it. Discovery is automatic; meaning stays human-controlled.

Governance Analysis

What policy would have decided, while staying in SHADOW.

Available now in Hosted Observa for evaluated activity. A dedicated walkthrough of these states is in production; until then, here is exactly what the words mean.

  • WOULD_ALLOWAn enforcing policy would have permitted this action.
  • WOULD_DENYIt would have refused — and the action still ran, because nothing is enforced.
  • WOULD_REQUIRE_APPROVALIt would have held this for a human. Nothing was held.

Counterfactual by construction. Not every runtime action receives a decision; activity never becomes a decision on its own, and unevaluated activity stays visible as UNEVALUATED.

How to read these states →
Explore the beta Read the docs
Private beta · self-service request, human-reviewed access

Now onboarding selected n8n and OpenClaw workflows.

Start the request yourself from the command line; a person reviews every one. Bring one bounded workflow, map it, capture governance evidence, and independently verify supported outcomes. SHADOW mode first. No production control, billing, or enforcement in this beta.

npm install -g @mcpherson-ai/observa-local-node
observa request-access --api-url https://governance-plane-observa.tailb473db.ts.net
Request Access Read the docs
The accountability problem

A successful runtime response is not proof the work happened.

An agent can report success while the downstream system tells a different story. A log entry can be useful without answering the questions an operator, reviewer, or incident owner actually needs resolved.

  • A completion may not proveThe correct action occurred.
  • A completion may not provePolicy was followed.
  • A completion may not proveApproval was required or obtained.
  • A completion may not proveThe intended downstream state exists.
  • A completion may not proveThe evidence is sufficient.
  • A completion may not proveThe issue is resolved.
What Observa does

Five capabilities that stay honest about their limits.

01

Discover

Identify the agents, tools, and capabilities operating in the connected environment.

02

Map

Generate reviewable mappings between capabilities and governed meaning. A person reviews and classifies before anything counts.

03

Observe

Capture structured execution evidence without taking control of production actions.

04

Govern

Evaluate mapped activity against defined governance boundaries in shadow mode.

05

Verify

For supported workflows and authoritative systems, independently determine whether the intended state actually exists.

How the governed path works

A readable path from action to verified reality.

Each stage has a distinct job. The current beta observes, records, and verifies; it does not give any policy decision authority over execution.

  1. Agent actsSupported activity enters the observed path.
  2. Observa observesAttempts and completions are recorded with minimized metadata.
  3. Policy evaluatesMapped activity is checked against explicit rules in shadow mode.
  4. Evidence is boundAttempt, decision, and observed completion are correlated.
  5. Reality is checkedSupported outcomes are compared with authoritative system state.
  6. Operator reviewsVerified results, mismatches, and unknowns stay distinct.
Current beta boundary

What this beta does, and what it does not

What it does

  • Discovers agents, tools, and capabilities on supported n8n and OpenClaw paths
  • Generates mappings for human review and classification
  • Captures structured, reviewable governance evidence
  • Evaluates policy in shadow mode
  • Independently verifies supported workflow outcomes against authoritative system state
  • Keeps mismatches and uncertainty visible instead of assuming success

What it does not do

  • Does not block, approve, deny, or rewrite agent actions
  • Does not take control of production systems
  • Does not treat a runtime completion as proof of the downstream state
  • Does not claim every workflow can be automatically verified
  • Does not certify compliance, completeness, or safety

“We would rather tell you what we cannot see than pretend we see everything.”

The public OpenClaw plugin

One public plugin. One private beta. Two different things.

The McPherson Governance OpenClaw plugin is free, public, and account-free, with its own verified release line. The Observa private beta is the broader product: n8n and OpenClaw workflows, founder-assisted mapping, governance evidence, and independent verification. The evidence below applies to the public v0.7.3 plugin release.

Reproducible archive verifiedThe published .sha256 asset equals a deterministic rebuild
Public GitHub releaseLatest v0.7.3 · annotated tag at 1237c59a
Public ClawHub package availableListed and installable · ClawHub security audit: Safe
Shadow-only authority boundary verifiedActive enforcement remains false
Install the Free Plugin View the public release Review proof and research →
Managed starting point

Founding Governance Setup

For OpenClaw operators who want help mapping agent activity, defining initial policy, reviewing evidence quality, and identifying coverage gaps before considering enforcement.

Request Founding Setup
Company lanes

Governance leads. Operating experience stays visible.

Governance

McPherson Governance

The public OpenClaw plugin, shadow policy evaluation, evidence binding, and the Observa governance and verification layer.

Explore Governance →

QSR Systems

Built from operations

Restaurant operating workflows developed from sixteen years of direct operating experience.

See QSR Systems →

Workflow Services

Narrow, accountable builds

Workflow mapping, source-of-truth discovery, approval design, regulated discovery, and narrow automation builds.

See Workflow Services →